#060 ML Егор Самосват. Монетизация, рекомендации и причём здесь ML
#060 ML Егор Самосват. Monetization, recommendations and what does ML have to do with it
Keywords
Summary
218 words
Critical Evaluation
Value of the Information & Strength of the Argument
The episode offers substantial value for listeners interested in the intersection of machine learning and business monetization. Egor Samosvat provides a clear conceptual framework: monetization as a balance between platform revenue and user value, the classification of promotion products by risk distribution, and the rationale for auctions. The argumentation is structured and logical, moving from definitions to mechanisms to ML implementation. The guest uses concrete examples, such as the devaluation of promotion when everyone buys it, and the difference between prepaid and postpaid systems in terms of measuring model improvements. He also addresses the user perspective, acknowledging banner blindness and the need to avoid degrading experience. However, the discussion remains largely qualitative; no specific metrics, A/B test results, or model architectures are shared. The reasoning is sound but relies on the guest’s authority rather than on publicly verifiable data. The argument for postpaid models is persuasive but one-sided, as potential downsides (e.g., revenue volatility, fraud) are not explored in depth.
Scientific Rigor, Source Quality, Title Accuracy
The episode does not cite academic papers or external studies directly. The description includes links to a lecture playlist on algorithmic game theory and a talk on monetization at Avito, which are relevant references. The guest’s expertise is evident, but the claims are not backed by data or citations within the conversation. The title accurately reflects the content, as the episode covers monetization, recommendations, and ML. The promotional segment for Practical ML Conf is clearly separated and does not interfere with the technical discussion. The host occasionally interjects with personal anecdotes, which adds context but also shifts focus away from rigorous analysis. Overall, the source quality is moderate: the information comes from a credible industry practitioner, but the lack of external validation limits its scientific rigor.
301 words
Title / Content Match
The title accurately reflects the content: the episode focuses on monetization mechanisms, recommendation-like ranking, and the role of machine learning, with Egor Samosvat as the guest.
Quality & Reliability
7/10
The guest is a credible practitioner leading the monetization efficiency unit at Avito, with a decade of ad-tech experience at Yandex. The discussion is grounded in real production systems (auctions, ranking, post-payment vs prepayment products) and avoids exaggerated claims. However, no quantitative results, datasets, or peer-reviewed references are provided, and the episode is partly promotional for a conference and podcast courses, which slightly lowers the overall reliability score.
Key Moments
Markers derived by PSI from the transcript: the creator did not define chapters.
- Introduction and promotional segment for Practical ML Conf.
- Egor Samosvat introduces himself and his role at Avito.
- Discussion on the elasticity of promotion services and the need for auctions.
- Definition of monetization and classification of promotion products by risk distribution.
- Comparison of prepaid vs postpaid models and their impact on measuring model improvements.
- How machine learning is used in monetization: similarities and differences with recommendation systems.
- Explanation of the auction mechanism for ad positions and how bids are converted.
- User experience considerations: banner blindness and the balance between revenue and trust.
- Discussion on long-term strategies and healthy greed benefiting all participants.
- Closing remarks and final thoughts.
Cited Sources
- Lecture playlist on Algorithmic Game Theory — Referenced in the description as a resource for understanding auction theory and game theory.
- Talks on monetization at Avito — Linked in the description as a relevant talk on monetization at Avito.
External References
Contribution & Novelties
The episode provides a practitioner’s perspective on how machine learning enables monetization at a large classifieds platform. It highlights the shift from simple prepaid promotion to auction-based, postpaid models, and explains why this alignment of incentives is crucial for measuring the impact of ML improvements. The discussion on the elasticity of promotion services and the role of auctions in dynamically balancing supply and demand is particularly insightful. The guest also draws a clear parallel between monetization and recommendation systems, emphasizing the auction as the key differentiator.
Pour aller plus loin :
- Auction theory — Wikipedia article covering the theoretical foundations of auctions, relevant to the discussion on ad auctions.
- Mechanism design — Wikipedia article on the field that studies how to design rules and incentives, closely related to the auction mechanisms described.
- Recommender system — Wikipedia article providing background on recommendation systems, which the guest compares to monetization ranking.
149 words
Radar Profile
The radar profile shows a strong emphasis on information quality and reliability, with moderate technical depth. The quantity of information is good but not exhaustive, and the technical level is accessible to a non-specialist audience. The overall profile reflects a well-structured expert discussion that balances conceptual clarity with practical insights.